Restorative vs Punitive: Two Approaches to a First-Time AI Violation